About

AI Coverage collects reporting on artificial intelligence from hundreds of publications, groups it by event rather than by article, and shows who covered each one — including who didn’t.

The problem

A single announcement can produce sixty articles across forty outlets in a day. Most read alike, because most are rewrites of the same source. A few are not: they add reporting, or they frame the event in a way the others don’t. Finding those few means reading all sixty.

Every story here is one real-world event with every article we found attached to it. The bar under each headline shows how coverage splits across camps that habitually disagree about AI, so you can see at a glance whether a story was reported broadly or only by people already inclined to care about it.

What each story carries

• A neutral summary written to adopt no outlet’s framing.
• The facts every outlet reports the same way.
• Where outlets differ — in wording, framing, or the figures cited.
• Every article, linked to its publisher, grouped by the stance of the piece.

What we don’t do

We don’t republish anyone’s work. Story pages link out to the original and always will. Article text is fetched only to analyse it and is discarded afterwards.

We don’t rate publications. Evidence grades describe how a claim is sourced — a paper, a shipped product, a press release, an anonymous source. A press release carried by an excellent newspaper is still a press release, and saying so is not a comment on the newspaper.

Who runs it

AI Coverage is built and run by Sam, working alone. It began as a personal tool: keeping up with AI meant reading the same event eight times in eight framings, and nothing on the market clubbed it together properly. It turned out to be worth showing other people.

AI Coverage is an independent company. It is not owned by, funded by, or affiliated with any company it covers, any publisher whose reporting appears here, or any other venture of its founder. There is no advertising and no sponsored placement.

Write to hello@aicoverage.org — corrections, source suggestions and complaints all reach Sam directly, and he is the person who answers them.

How this is made, and who is accountable for it

The pipeline is automated and the analysis is done by a language model. We would rather say that plainly than have you work it out. Here is the division of labour:

• The software does: collecting articles, extracting what happened from each one, grouping articles into events, detecting syndicated reprints, and drafting the summary and the agreement and disagreement lists.
• Sam does: deciding what the system measures and how, reviewing grouping and summary quality, correcting what readers report, and taking responsibility when it is wrong.

The method is documented in full on the method page, including what the pipeline gets wrong. No summary here is presented as original reporting: every claim belongs to an outlet, and that outlet is named and linked on the story page. If a summary misrepresents a source, that is our error and not theirs — tell us and we will fix it. See corrections.